Observed arrival · 2026-11-12
Real Unreal Bench: AI on real Unreal Engine tasks
A benchmark report tests eight models on 15 game-development tasks across nine Unreal Engine games, scoring whether the running games pass deterministic checks.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Unreal Engine developers evaluating coding models
- Worth noticing
- Each model gets one run per task; the site gives 95% Wilson intervals and notes they do not account for shared games or task selection.
Field notes
The results are divided into overall performance and task categories, with a cost-versus-pass-rate comparison alongside the leaderboard. The page reports mean model cost per attempted task from $2.93 to $28.16, excluding virtual machines. It also flags two limits on interpreting the intervals: tasks share games, and selection favored cases that earlier Codex runs found difficult.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
ƒJavaScriptBrowser-side code central
One card from the complete issue